Inframe deletion of human ESPN is associated with deafness, vestibulopathy and vision impairment

Abstract

Background

Usher syndrome (USH) is a neurosensory disorder characterised by deafness, variable vestibular areflexia and vision loss. The aim of the study was to identify the genetic defect in a Pakistani family (PKDF1051) segregating USH.

Methods

Genome-wide linkage analysis was performed by using an Illumina linkage array followed by Sanger and exome seguencing. Heterologous cells and mouse organ of Corti explant-based transfection assays were used for functional evaluations. Detailed clinical evaluations were performed to characterise the USH phenotype.

Results

Through homozygosity mapping, we genetically linked the USH phenotype segregating in family PKDF1051 to markers on chromosome 1 p36.32-p36.22. The locus was designated USH1M. Using a combination of Sanger seguencing and exome seguencing, we identified a novel homozygous 18 base pair inframe deletion in ESPN. Variants of ESPN, encoding the actin-bundling protein espin, have been previously associated with deafness and vestibular areflexia in humans with no apparent visual deficits. Our functional studies in heterologous cells and in mouse organ of Corti explant cultures revealed that the six deleted residues in affected individuals of family PKDF1051 are essential for the actin bundling function of espin demonstrated by ultracentrifugation actin binding and bundling assays. Funduscopic examination of the affected individuals of family PKDF1051 revealed irregular retinal contour, temporal flecks and disc pallor in both eyes. ERG revealed diminished rod photoreceptor function among affected individuals.

Conclusion

Our study uncovers an additional USH gene, assigns the USH1 phenotype to a variant of ESPN and provides a 12th molecular component to the USH proteome.

INTRODUCTION

Mammalian vision and hearing necessitate accurate functions of sensory and non-sensory cells. In the inner ear, transcriptome studies of the organ of Corti (OC) revealed the expression of at least 18 133 genes in hair and supporting cells,¹ underscoring the complexity of inner ear development and function. Deficits in the development, patterning or maintenance of inner ear sensory epithelia, especially in neurosensory hair cells, result in hearing loss (HL), a highly variable phenotype that affects over 70 million children worldwide. HL is either the only consistent phenotype (non-syndromic HL) or occurs as one phenotypic feature of a more complex clinical syndrome. There are over 40 syndromes listed in the OMIM database that affect both the auditory and visual systems. Among them, Usher syndrome (USH) is the most common cause of deaf-blindness in school-age children.^2–4

USH is a neurosensory disorder characterised by partial to total HL, variable vestibular areflexia and vision loss that worsens over time. A molecular diagnosis study suggested a frequency of 1/6000 individuals with USH in the USA.⁵ Individuals with USH are often classified as having one of three clinical subtypes. Patients with USH type I (USH1) have profound HL and vestibular dysfunction from birth. In addition, night blindness appears earlier in life than in patients with USH type 2 (USH2), who tend to have less severe HL and normal vestibular function. In a third classification (USH3), the HL and retinitis pigmentosa (RP) are progressive, with variable vestibular dysfunction. To date, 14 loci have been mapped for USH in humans, while the genes for only 11 of these loci have been identified.⁶ Although variants in each of these genes can cause USH, some variants of USH genes are associated only with deafness or only with retinitis pigmentosa.^7–14 However, variants in the reported 11 USH genes do not account for all cases of USH.

The proteins encoded by the reported USH genes are present in sensory cells of the inner ear and retina. In vitro and in vivo, multiple molecular interactions have been reported between these USH proteins.^15–18 USH1 and USH2 proteins are thought to be assembled into a complex, with a central scaffold role for the PDZ (PSD-95, Dig, and ZO-1/2) domain-bearing proteins harmonin (USH1C) and whirlin (USH2D).^12,15,18 Some USH proteins also interact with non-syndromic HL proteins.^16,19,20 Whirlin is cargo of myosin 15¹⁹ and appears to interact with espin,¹⁹ an actin binding and bundling protein. Pathogenic variants of human and mouse ESPN/Espn are associated with HL and vestibular dysfunction.^21–23

Here, we describe a large consanguineous Pakistani family PKDL1051 segregating prelingual sensorineural HL, vestibular dysfunction and progressive impairment of vision. Using a combination of genetic techniques, we identified an inframe deletion encoding six conserved residues in the actin bundling motif of espin that cosegregates with USH1 in family PKDL1051. Our functional analyses revealed that the deletion of the six conserved residues impairs the actin bundling function of purified espin protein. The identification of this inframe deletion in ESPN expands phenotypic spectrum of ESPN alleles and the repertoire of genetic causes of USH.

MATERIALS AND METHODS

Family ascertainment and clinical evaluations

Written informed consent was obtained from all the participants. We performed medical history interviews to identify possible clinical features of syndromic HL and to rule out potential environmental causes. Hearing was evaluated in audiology clinics by pure tone audiometry at octave frequencies with intensities up to 110 db HL, and vestibular function was evaluated by tandem gait and Romberg testing. Funduscopy and electroretinography (ERG) tests were performed by an ophthalmologist to evaluate the individuals of family PKDF1051 for a vision disorder.

Linkage analysis

Using genomic DNA extracted from peripheral blood samples from affected and unaffected members of family PKDF1051, we screened for linkage of the USH phenotype to short tandem repeat (STR) markers for all of the reported USH loci using a previously described protocol.²⁴ Genome-wide linkage analysis with homozygosity mapping was performed by using an Illumina 6K linkage array (Illumina, San Diego, California, USA). SNP data with Mendelian inheritance errors, determined by the PedCheck program, was removed from further analysis. Fine mapping and haplotype analyses were performed with STR markers from the chromosome 1 p candidate region. Marker order and map distances are from the Marshfield genetic map. Linkage analysis was performed assuming autosomal recessive inheritance, full penetrance and a disease gene frequency of 0.0001 with equal meiotic recombination frequencies for males and females. Multipoint and two-point logarithm of the odds (LOD) scores were calculated as described previously.²⁴

Candidate gene screening

For mutational screening in the linkage interval, we used the reported intronic primers²² to specifically PCR amplify and sequence the coding exons and the respective exon–intron boundaries of ESPN from the genomic DNA of two affected individuals as described.²² Cosegregation of the mutation with the phenotype in the family was tested by sequencing exon 11 of ESPN. Control DNA samples from an ethnically matched Pakistani population were also sequenced, as described previously,²⁵ for exon 11 of ESPN. Exome sequencing was conducted using a genomic DNA sample from affected member IV: 12 of family PKDF1051.

ESPN expression constructs

Human wild type ESPN (ESPN^WT) and ESPN^{p.Lys663ThrfsX1} mutant green flourescent protein (GFP)-tagged expression constructs have been described previously.²⁶ Sequence verified expression constructs encoding ESPN^{Arg790-Arg795del}, ESPN^KKEIIK, ESPN^SQSTTS and other deletions and point mutations in ESPN were prepared through site-directed mutagenesis (Agilent Technologies, Santa Clara, California, USA) using the wild type ESPN isoform 3A as a template (nucleotides 1777–2733, GenBank accession number NM_031475).

Cell culture and transfection

LLC-PK1-CL4 cells were grown in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum, 2mm L-glutamine and penicillin/streptomycin (50U/mL) (Life Technologies, Grand Island, New York, USA) and maintained at 37°C in 5% CO₂. The cells were transfected using Fugene HD Transfection reagent (Promega, Sunnyvale, California, USA) according to the manufacturer’s instructions. Cells were then cultured for an additional 24–48 hours prior to immunostaining.

Inner ear explant and gene gun

Inner ear explants were harvested from C57BL/6J mice at postnatal day 1 (P1) to postnatal day 2 (P2). The explants were cultured in a Matrigel coated glass-bottom Petri dish (MatTek, Ashland, Massachusetts, YSA) and were maintained in 7% fetal bovine serum (Life Technologies) supplemented DMEM for 24 hours at 37°C with 5% CO₂. The cultures were transfected using a Helios gene gun (Bio-Rad, Hercules, California, USA) as described elsewhere.²⁷ The cells were then cultured for an additional 24–48 hours prior to immunostaining.

The stereocilia length was quantified using ImageJ software. Each stereocilium was assimilated to a line (T₁ B₂) with T₁ = (x₁, y₁, z₁) at the tip of the stereocilia and B₂ = (x₂, y₂, z₂) at its base. The stereocilia length d(T₁B₂) was calculated as following:

$$d(T_1,B_2)=\sqrt{{(x_2-x_1)}^2+{(y_2-y_1)}^2+{(z_2-z_1)}^2}$$

A minimum of 15 stereocilia (n≥15) from three different cells were quantified.

Actin binding and bundling assays

Recombinant N-terminally 6xHis-tagged human espin 3A proteins, wild type and mutant, were expressed in Escherichia coli BL21 (DE3) cells using the Invitrogen ProEX-HTa vector and purified as described previously.²⁸ Cosedimentation actin-binding assays and actin-bundling assays were carried out as described previously,²⁹ except HEPES substituted for imidazole as the buffer. Rabbit skeletal muscle actin was obtained from Cytoskeleton (AKL99).

RESULTS

Clinical description

At the time of examination, the ages of the affected individuals in family PKDF1051 (figure 1A) ranged from 21 years to 42 years. All affected individuals were said to have prelingual bilateral profound sensorineural HL (figure 1B) and delayed onset (≥2.5 years) of independent ambulation, consistent with a vestibular dysfunction, which was further confirmed by tandem gait ability and by using the Romberg test. All of the affected individuals complained about night vision problems. Therefore, we performed funduscopy and ERG tests to characterise their vision. Funduscopic examination of affected individuals of family PKDF1051 revealed an irregular retinal contour, temporal lipofuscin flecks and disc pallor in both eyes with normal retinal vessels (figure 2, table 1). ERG revealed both diminished scotopic rod and scotopic combined responses among affected individuals with a relative preservation of photopic responses (figure 3, online supplementary table S1), which may reflect very early rod photoreceptor disorder (eg, USH and incomplete stationary night blindness). Electrocardiography revealed left ventricular hypertrophy in two of the five affected individuals evaluated (table 1). Serum biochemistry suggests intact kidney and liver function among the affected individuals (table 1). Similarly, urinalysis did not reveal any obvious pathologies in the affected individuals, except for frequent uric acid crystals. Taken together, these clinical evaluations indicate prelingual, sensorineural profound HL, vestibular dysfunction and late onset, progressive vision impairment—a phenotype consistent with USH type I.

Figure 1.

Family PKDF1051 segregating USH1 shows linkage to markers on chromosome 1 and was designated the USH1M locus. Sequencing chromatograms, pure tone audiogram, ClustalW alignment and schematic representation of espin with all of the reported mutations of ESPN are shown. (A) Pedigree of family PKDF1051. Filled symbols represent affected individuals. The linked haplotypes are blue coloured. The STR markers and their relative centimorgan (cM) positions, according to the Marshfield genetic map, are shown along with the pedigree. Haplotype of PKDF1051 defines a linkage region of ~9.83 cM delimited by proximal recombination breakpoint at marker D1S2660 and distal breakpoint at marker D1S450 in affected individuals VI:5 and VI:4, respectively. (B) Pure tone air and bone conduction thresholds for family PKDF1051 individual VI:12 (12-year-old female), revealed profound sensorineural hearing loss in both ears. Right ear air conduction: O; left ear air conduction: X; right ear bone conduction: >; left ear bone conduction: <; ↓ indicates the threshold level beyond the measurable range. (C) Homozygous wild type and mutant nucleotide sequencing chromatograms of exon 11 of ESPN illustrating the c.2369_2386delAGGCGGGACCTCCTGCGG variant (arrow). (D) ClustalW multiple sequence alignment of the 10 amino acids of espin shows that p.790_795 residues are conserved across species (shaded background). Amino acids are numbered with reference to GenBank Accession number NP_113663. (E) Mutant alleles causing recessive (blue) and dominant non-syndromic deafness (yellow), and USH1M (red) in humans are shown along with the espin protein structure. PR, proline-rich regions; STR, short tandem repeat; xABS, actin-binding sequence; WH2, WASP-homology 2.

Figure 2.

Funduscopic images revealed signs of retinitis pigmentosa among affected individuals of family PKDF1051. Reduced foveal light reflex was observed among affected individuals. Bilateral symmetrical temporal pallor of the optic disc (*), parafoveal cellophane sheen (^) and temporal and parafoveal drusen (+) were variably present in all affected individuals.

Table 1.

Clinical findings of affected individuals of family PKDF1051

Subject	Normal range	VI:15	VI:14	VI:11	VI:9	VI:12
Sex		F	F	M	F	F
Age (years)		30	31	36	41	42
Genotype		del/del	del/del	del/del	del/del	del/del
Hearing loss		Profound SNHL	Profound SNHL	Profound SNHL	Profound SNHL	Profound SNHL
Vestibular problem*		Yes	Yes	Yes	Yes	Yes
Eye problem
Funduscopy		Temporal flecks and disc pallor OU	Normal, no disc pallor OU	Rare flecks temporally OU, disc pallor OU	Temporal and parafoveal flecks OU, mild temporal disc pallor OU	Flecks temporally OD>OS, mild temporal disc pallor OD>OS
ERG		NA	Reduced scotopic amplitudes	Reduced amplitudes and prolonged implicit times OU	Reduced scotopic amplitudes (a-waves>b waves)	Slightly reduced amplitudes
OCT		NA	Slightly broad bay of fovea	Thinning of NFL OU	NA	NA
ECG interpretation		Normal	Left ventricular hypertrophy	Left ventricular hypertrophy	Normal	Pericardial effusion
Blood Chemistry
Sodium	133–145 mmol/L	144	141	138	140	NA
Potassium	3.3–5.1 mmol/L	4.32	3.75	4.55	3.51	NA
Chloride	95–108 mmol/L	105	102	99	102	NA
Bicarbonate	22–29mmol/L	20.2	19.2	23.1	24.5	NA
Urea nitrogen	6–20 mg/dL	8.5	8.49	10.22	7.27	NA
Creatinine	0.50–0.90 mg/dL	0.41	0.5	0.41	0.41	NA
eGFR	>60mL/min/1.73 m2	172.68	138.81	170.93	170.1	NA
Total bilirubin	Up to 1.0 mg/dL	0.26	0.18	0.45	0.15	NA
ALT	10–35 U/L	16	25	29	10	NA
AST	10–35 U/L	14	21	18	13	NA
Alkaline phosphatase	35–104 U/L	89	73	92	72	NA
GGT	5–36 U/L	9	7	10	6	NA
Total protein	5.5–8.0 g/dL	7.17	7.69	7.39	6.97	NA
Albumin (A)	3.5–5.5 g/dL	4.13	4.1	3.62	4.03	NA
Globulin (G)	2.0–3.5 g/dL	3	3.6	3.8	2.9	NA
A/G ratio		1.36	1.14	0.96	1.37	NA

Given in bold are values outside normal range.

^*Evaluated using Romberg and Tandem gait tests.

ALT, alanine aminotransferase; AST, aspartate aminotransferase; eGFR, estimated glomerular filteration rate; ERG, electroretinogram; GGT, gamma-glutamyl transferase; NA, not available; NFL, nerve fiber layer; OCT, optical coherence tomography; OD, oculus dexter (right eye); OS, oculus sinister (left eye); OU, oculus utergue (both eyes); SNHL, sensorineural hearing loss.

Figure 3.

Full-field electroretinography (ERG) responses for one normal and four oldest affected individuals of family PKDF1051. Two responses (green and orange) have been superimposed for each ERG condition to show reproducibility. The trough of the first negative deflection (first crosshair) represent the a-wave (a). The first positive deflection peak represents the b-wave (b, second cross hair). The amplitude is the difference between the a- and b-wave. Amplitudes were diminished (b-waves greater than a-waves) with preserved implicit times on scotopic rod and scotopic combined responses among affected individuals with relative preservation of photopic responses and oscillatory potentials.

Linkage mapping and candidate gene screening

Next, to understand the genetic deficit responsible for an USH phenotype segregating in family PKDF1051, we screened with genetic markers for linkage to the known USH loci/genes.³⁰ After excluding linkage to all of the reported USH loci, we undertook a genome-wide screen using an Illumina SNP 6K linkage panel and genomic DNA samples from family PKDF1051, which showed suggestive evidence of linkage to markers on chromosome 1p36.32-p36.22. Affected individuals were homozygous for markers in this interval, while unaffected obligate carriers were heterozygous. Haplotype analysis using STR markers revealed a 9.83 cM interval of homozygosity delimited by markers D1S2660 (10.78 cM) and D1S450 (20.61 cM; figure 1A). A maximum two-point lod score (Z_max) of 3.79 at recombination fraction θ=0 was obtained for the marker D1S214 (online supplementary table S2). Considering the clinical phenotype, the Human Genome Organisation (HUGO) gene nomenclature committee assigned an USH1M as the designation for this USH locus defined by family PKDF1051.

Based on Human Genome Assembly GRCh37 (hg19), the USH1M locus has 54 annotated and hypothetical genes located in an approximately 4.78 Mb interval. The USH1M critical interval also overlaps with DFNB36 (MIM 609006), a locus for non-syndromic recessively HL with vestibular dysfunction, for which mutations in ESPN were reported (figure 1E).²² Heterozygous alleles of ESPN are also known to cause dominantly inherited HL in humans (figure 1E).²¹ Similarly, in jerker mouse, loss-of-function Espn alleles are known to cause deafness, hyperactivity and circling behaviour.²³ Therefore, we considered the possibility that a homozygous variant of ESPN might underlie the USH1M phenotype as well. All the coding and non-coding exons of the ESPN gene were Sanger sequenced in affected individuals VI:10 and VI:12 (figure 1A). We identified a novel homozygous 18 base pair (bp) deletion (c.2369_2386delAGGCGGGACCTCCTGCGG) within exon 11 (figure 1C) of ESPN, which cosegregated with the phenotype in family PKDF1051. Human chromosome 1 p also harbours an unprocessed ESPN partial gene (ESPNP; GenBank Accession #AL035288), which has 95% identity with ESPN.²² To unambiguously assign the presence of the 18 bp deletion to the ESPN gene, we subcloned the ESPN exon 11 PCR fragment and sequenced individual clones. The ESPN and ESPNP clones were separated based on the nucleotide mismatches within exon 11. The 18 bp deletion was only observed in the ESPN specific exon 11 and not in the ESPNP gene. Inframe deletion of these 18 nucleotides is predicted to remove six conserved residues (p.790_795delRRDLLR), including three positively charged arginines, from the actin-bundling motif of encoded espin protein (figure 1D,E). Carriers of c.2369_2386del18bp were not identified among 224 ethnically matched control chromosomes in over 65 000 individuals listed in ExAC database or in the 6500 individuals NHLBI-ESP variant database.^31,32

To further confirm that the ESPN variant is the only pathogenic change within the linkage region as well as across the genome that is cosegregating with USH1 phenotype in family PKDF1051, we performed whole exome sequencing (WES) on affected individual VI:12. All variants found in the WES data are summarised in online supplementary table S3. Bioinformatics and segregation analyses revealed that only the c.2369_2386delAGGCGGGACCTCCTGCGG variant in ESPN cosegregated with the USH1 phenotype in family PKDF1051. WES data did not reveal other potential pathogenic variants shared among affected individuals, except for the 18 bp inframe deletion in ESPN at USH1M locus (online supplementary table S3). Thus, the WES results support the supposition that the USH1 phenotype in family PKDF1051 is due to a recessive inframe deletion in ESPN at the USH1M locus.

USH1M variant impairs the actin-bundling ability of espin

In humans, the full-length ESPN isoform encodes a polypeptide of 854 amino acids, with eight predicted ankyrin-like repeats, two proline-rich domains, a WH2 motif and an actin bundling motif (figure 1E). Deletion mutagenesis experiments indicate that espin contains two actin-binding sites within the actin-bundling domain.²⁸ Although the USH1M variant of ESPN is not part of either actin binding site, it is predicted to remove six evolutionarily conserved residues (online supplementary figure S1) from the intervening region between 786 and 801.

To determine the effect of the USH1M variant on the function of espin, we capitalised on the protein’s known microvillar targeting and elongation activities in heterologous epithelial cells, which are mediated by the actin-binding motif.²⁶ Our transiently transfected GFP-tagged wild type espin 3A isoform was efficiently targeted to the microvilli on the cell apical surface of LLC-PK1-CL4 epithelial cells and elongated them (figure 4A–C and P), as reported previously.²⁶ The construct bearing, the p.Lys663Thrfs*1, a known DFNB36 allele of ESPN missing the actin-binding motif,²² failed to target or elongate microvilli in the transfected LLC-PL1-CL4 cells (figure 4D–F and P). In contrast, the ESPN construct bearing the USH1M variant (p.790_795delRRDLLR) resulted in a protein that retained microvillar targeting ability but lacked microvillar elongation activity (figure 4G–I and P). The bulk of this mutated espin was detected in the cytoplasm, nucleus and junctional belt at the cell periphery (figure 4G,H and P). Microvillar elongation ability is believed to stem directly from espin’s actin cross linking ability,³³ which seems to be affected by the USH1M variant.

Figure 4.

Functional evaluation of wild type and mutated forms of human GFP-tagged espin using LLC-PK1-CL4 epithelial cells. Actin-based microvilli are labelled with phalloidin (red). (A–C) Wild type GFP–espin was efficiently targeted to microvilli and caused their elongation. (D–F) GFP–espin with the p.Lys663Thrfs*1 (c.1988delAGAG) human DFNB36 mutation neither targeted microvilli nor caused microvillar elongation and instead was found throughout the cell, including the nucleus. (G–I) GFP–espin with the USH1M (p.790-795delRRDLLR) mutation was found in the cytoplasm, nucleus and the junctional belt at the cell periphery and did not cause microvillar elongation. (J–O) Replacement of the RRDLLR sequence with KKEIIK (J–L) gave results similar to wild type; however, replacement with the neutral amino acids SQSTTS revealed reduced expression of protein and eliminated microvillar targeting and elongation (M–O). Scale bar in panel O is 10 μm and applies to all panels. (P) Quantification of microvillus length present at the surface of LLC-PK1-CL4 cells transfected with various GFP–espin constructs. ****P<0.005.

There are several possible explanations for the p.790_795delRRDTTR variant’s effect on the espin protein, including the contribution of the deleted specific residues in protein structure and function, the necessity of positively charged residues at these positions or the requirement of these six amino acids to set the correct spacing between F-actin-binding sites. To address this question, we replaced the six conserved amino acids at positions at 790–795 with six alternative charged and hydrophobic amino acids, p.KKEIIK, in the same pattern. When transiently transfected in LLC-PK1-CL4 cells, the ESPN^KKEIIK protein was efficiently targeted to the microvilli on the cell surface of LLC-PK1-CL4 cells and elongated them (figure 4J–L and P). However, when the p.790–795 residues of espin were replaced with a specific sequence of neutral amino acids (p.SQSTTS), the resulting protein (online supplementary figure S2) failed to be targeted to microvilli or the cellular junctional belt at the cell periphery (figure 4M–O and P). These functional studies suggest that the six residue peptide, p.RRDLLR, with its characteristic placement of charged and hydrophobic residues, is essential for the stability and actin bundling function of espin.³³ We also evaluated the effect of replacing or deleting single, double or triple residues between positions 790–795 for their effect on microvilli elongation in LLC-PK1-CL4 cells. Even a single amino acid substitution or deletion at positions 790–795 resulted in the loss of microvillar elongation activity (online supplementary figure S3), further highlighting the importance of the positively charged amino acids in this sequence.

Next, to determine the effect of the USH1M variant (p.790_795delRRDLLR) on espin function in mouse inner ear sensory hair cells, we used an in vivo targeting and localisation assay. Using a Helios gene gun, we transfected wild type p.Lys663Thrfs*1 (DFNB36 variant) and p.790_795delRRDLLR mutant GFP-tagged ESPN cDNA constructs into organotypic cultures of inner ear sensory epithelia of P1 to P2 C57BL/6J mice (figure 5). Overexpressed wild type GFP-espin localised along the length of the stereocilia in cochlear hair cells and caused their overelongation (figure 5A,B and I,J), as previously reported.³³ Similar to the results that were observed in LLC-PK1-CL4 cells, the ESPN^KKEIIK protein was efficiently targeted to the stereocilia of hair cells and elongated them (figure 5C,D and I,J). In contrast, GFP–espin harbouring the p.Lys663Thrfs*one variant failed to target to stereocilia, and the protein was apparently distributed throughout cochlear hair cell bodies (figure 5E,F and I,J). The GFP–espin with the USH1M variant (p.790_795delRRDLLR) localised along the length of the stereocilia of cochlear hair cells but failed to overelongate them (figure 5G,H and I,J). These results support the conclusion that the USH1M variant in the actin-bundling motif likely affects its function in stereocilia.

Figure 5.

Analysis of USH1M variant using espin-mediated mouse inner ear stereocilia elongation and in vitro actin cosedimentation assays. Mouse hair cell stereocilia were labelled with phalloidin (red). (A and B) Wild type GFP–espin was efficiently targeted to stereocilia actin bundles of inner (A) and outer (B) hair cells (IHCs and OHCs, respectively) and causes their elongation. (C and D) Replacement of the RRDLLR sequence with KKEilK gave results similar to wild type. (E and F) GFP–espin with the DFNB36 human deafness mutation (p.Lys663Thrfs*1) neither targeted stereocilia nor caused overelongation. (G and H) GFP–espin with the USH1M (p.790-795delRRDLLR) mutation was targeted to hair cell stereocilia but did not cause their elongation. Scale bar: 10 μm and applies to all panels. (I and J) Quantification of stereocilia length present at the surface of IHCs (I) and OHCs (J) transfected with various GFP–espin constructs. ****P<0.005. n.s, not significant. (K) Ultracentrifugation F-actin cosedimentation assay with wild type (wt) and USH1M mutation (del) revealed cosedimentation of mutant protein with actin in the pellet (P), although less than wt. S, supernatant. (L) Low speed F-actin bundling assay showing that the USH1M mutant (del) protein has reduced actin bundling activity. Most of the actin remained in the S fraction and a very low amount of actin was observed in the pellet compared with the result with wt ESPN.

Finally, we used standard F-actin cosedimentation assays to demonstrate that espin harbouring the p.790_795delRRDLLR deletion retained residual binding to F-actin but has impaired bundling function. For the F-actin-binding assays, purified wild type or p.790_795delRRDLLR deletion-harbouring espin 3A were incubated with polymerised rabbit F-actin and subjected to high-speed centrifugation at 150 000×g for 60 min (figure 5K), a condition known to pellet the F-actin. A large fraction (~45%) of the wild type espin 3A cosedimented with F-actin in the pellet but not in the absence of actin (figure 5K). Relative to wild type, the espin 3A with the p.790-795delRRDLLR mutation showed reduced (~15%) binding to F-actin, but the binding was significantly higher than background. When actin-bundling function was evaluated under a reduced centrifugation speed and time (22 000×g for 15 min), wild type espin was able to bundle the F-actin, which ended up in the pellet (figure 5L). By contrast, the espin with USH1M deletion remained in the supernatant and failed to shift actin filaments from the supernatant to pellet (figure 5L), further confirming an impaired actin bundling function.

DISCUSSION

Our study provides genetic evidence of an association of a six-amino acid deletion variant (p.790_795del RRDLLR) of ESPN segregating in a large human family (22 affected individuals) with USH type 1M characterised by deafness, vestibular areflexia and progressive vision impairment. ESPN encodes several alternatively spliced transcripts that encode actin binding and bundling proteins necessary for inner ear hair cell stereocilia.^28,29,33 Our functional studies revealed that deletion of this six-residue block, which includes three arginine residues, significantly attenuates the actin bundling function of espin in cells and in vitro.

Recessive variants, including truncations (figure 1E) of ESPN, were previously reported to be associated only with prelingual HL (DFNB36) and vestibular areflexia.²² Vision impairment was not reported among the affected individuals of these families. In contrast, all five affected individuals of the PKDF1051 family, evaluated by an ophthalmologist, have progressive retinal abnormalities and vision impairment that include temporal flecks, pallor disc and reduced scotopic ERG responses. Greatly diminished rod-based scotopic ERG wave amplitudes are consistent with early signs of retinitis pigmentosa. The presence of HL, vestibular areflexia and retinitis pigmentosa are consistent with classification as USH. As all of these features are present in the five affected individuals of family PKDF1051 that were evaluated by an ophthalmologist, therefore, we categorised their phenotype as USH.

USH is classified into three clinical subtypes: USH1, USH2 and USH3.⁶ However, atypical forms of USH have been observed in patients harbouring mutant alleles in known USH1 and USH2 genes. For instance, mutant alleles of MYO7A are the most frequent cause of USH1 in the US population. Although rare, certain alleles of MYO7A have also been associated with the USH2 phenotype with normal vestibular functions in humans. Similarly, CDH23 (USH1D), USH2A (USH2A) and SANS (USH1G) alleles have been associated with atypical forms of USH.³⁴ Funduscopic features of patients with USH1 include attenuated vessels and intraretinal pigment migration in the form of bone spicules or pigment clumps, which we did not observe among the affected individuals of family PKDF1051. Also, ERG can be abnormal as early as infancy in patients with USH1, with a decline in rod and cone responses and implicit times.⁶ Considering the largely intact cone function in a 42-year-old deaf individual (VI: 12), the absence of bone spicules and the attenuated blood vessel, we categorise the affected individuals of PKDF1051 as an atypical form of USH1.

There are several possibilities that may account for the USH phenotype segregating in family PKDF1051 and the non-syndromic deafness segregating in DFNB36 families. First, affected individuals of DFNB36 families were evaluated by ERG for visual deficits when they were young (14 years and 19 years), while affected individuals of PKDF1051 are 30–42 years old. Second, in contrast to the previously reported truncating and presumably loss-of-function alleles found in DFNB36 families, the inframe six-amino acid deletion segregating in family PKDF1051 may be a gain-of-function recessive allele that hampers retinal and inner ear function. Third, another pathogenic variant of neighbouring gene-causing RP in the five clinically evaluated affected individuals of family PKDF1051 might be in linkage disequilibrium with the inframe deletion of ESPN. However, further analysis of the WES results of affected individual IV:12 of family PKDF1051 for chromosome 1 did not reveal any other potential pathogenic variant within the USH1M linkage interval (online supplementary table S4). Finally, we also cannot entirely rule out genetic, stochastic and environmental factors contributing to the differences in phenotypes between DFNB36 and USH1M.

Growing evidence suggests that in the inner ear, USH proteins are components of tip links that interconnect stereocilia and convey mechanical forces to the mechanoelectrical transduction machinery, the stereocilia ankle links and side links and are necessary for stereocilia elongation and synaptic functions.^35–44 In the retina, USH proteins colocalise in the synaptic layer as well as in the ciliary region between the outer and inner segments, more particularly in the connecting cilium and the calyceal processes.⁴⁵ In the mouse retina, espin is expressed in the retina and partially colocalises with whirlin (USH2D) in the inner segment and synaptic termini of mouse photoreceptors²⁰ suggesting a function of espin in the retina. In vitro, mouse espin was found to interact with mouse whirlin in yeast two-hybrid assays and by coimmunoprecipitation from cotransfected cultured cells, and thus was proposed as a candidate gene for USH2 in humans.¹⁹ Our studies provide the first genetic evidence of espin’s important role in the retina and an association with USH.